Tracking Cities, 2011-2012

I am using the “CITY” identifier variable in the ACS. I’ve noticed that for several cities, I can follow them from 2005-2011 OR 2012 and later years. In other words, there are several cities that I cannot follow across these two time periods. Examples include Fresno, CA (city id 2370), Gary, IN (2470), or Las Vegas, NV (3490).

Can someone explain why this is and if there is any way to create a longer panel for these cities?

Thanks very much,

Dear Rhiannon,

Leading up to the 2010 decennial census, the Census Bureau changed the rules for delineating Public Use Microdata Areas (PUMAs). In prior censuses, the Bureau privileged city boundaries when delineating PUMAs. For the 2010-vintage (and 2020-vintage) PUMAs, the Bureau privileged census tract and county boundaries when delineating PUMAs. This change yielded PUMA boundaries that don’t align as well with cities because census tracts don’t always follow city boundaries.

When we create our CITY variable, we examine the relationship between PUMAs and cities (technically places in census terminology). For 2012-forward (which was the first ACS that use the 2010 PUMA boundaries), we identified cities with a mismatch error (available in the CITYERR variable) of 10% or less.

We also provide crosswalks describing the errors between PUMAs and all cities of > 75,000 people in these Excel spreadsheets - PUMA-city crosswalk and PUMA Match Summary by Place. You could use those Excel spreadsheets to create a longer panel, but you’ll do so by accepting larger errors of omission and commission.

Dave Van Riper

Hello Dave,

Thank you so much for your quick response. Can you help me understand a bit better how I would go about making a cross walk? For instance, Las Vegas (city id 2490) is an identifiable CITY from 2005-2011. If I wanted to track Las Vegas in years after 2011, what is the “new” city name that I would use based on the crosswalks?

Dear Rhiannon,

I would start with the “PUMA Match Summary by Place” file for 2010. This file has a record for Las Vegas city, and it provides the following:

  • State FIPS code (32)
  • IPUMS city code (3490)
  • List of 2010 PUMAs that intersect with 2010 Las Vegas city boundary (there were 7 in 2010)
  • List of 2010 PUMAs with a majority of their population inside the 2010 Las Vegas city limits (this is in the “Best-Matching PUMAs” column) (there were 5 in 2010 - 00401, 00402, 00405, 00406, 00407)
  • Sum of Errors (11.7758% for Las Vegas city in 2010 - it’s close to the 10% threshold we used for CITY assignment)

From this file, I would extract the State FIPS code, IPUMS city code, and the five PUMAs from the “Best-Matching PUMAs” column. Once you’ve extracted that information, you will need to restructure the data to look like:

state_fips ipums_city puma
32 3490 00401
32 3490 00402
32 3490 00405
32 3490 00406
32 3490 00407

You could then match this crosswalk to ACS samples from 2012-2021 by state_fips and PUMA. This will assign the IPUMS_city code to records in those PUMAs in each 2012-2021 sample. That extends your panel through 2021, albeit for a group of PUMAs that don’t quite match the Las Vegas city boundary.

You would repeat the exercise with the “PUMA Match Summary by Place” file for 2020. In that file, these 5 2020 PUMAs are the best match with 2020 Las Vegas city - 00401, 00404, 00405, 00406, 00407.

I hope this helps!